Towards Modeling and Simulation of Particulate Interactions with - - PowerPoint PPT Presentation

Oliver M. F. Browne Postdoctoral Research Associate Mechanical Engineering, University of Kentucky, Lexington, USA NCSA Blue Waters Symposium for Petascale Science and Beyond Sunriver Resort in Sunriver, Oregon, June 3rd 2019

Towards Modeling and Simulation of Particulate Interactions with High-Speed Transitional Boundary- Layer Flows

Collaborators and Funding

• Funding provided by Office of Naval Research under contract N00014-19-1-2223 with Dr.

Eric Marineau as program manager is gratefully acknowledged. PI for this project: Dr. Christoph Brehm (University of Kentucky),

• External collaborators: Prof. Hermann Fasel (University of Arizona), Anthony Haas

(University of Arizona), fruitful discussions on particle modeling with Prof. Anatoli Tumin (University of Arizona)

• This research is part of the Blue Waters sustained-petascale computing project

, which is supported by the National Science Foundation (awards OCI-0725070 and ACI- 1238993) and the state of Illinois. Blue Waters is a joint effort of the University of Illinois at Urbana-Champaign and its National Center for Supercomputing Applications,

• An extended form of this presentation will be given at AIAA Aviation conference in Dallas,

Texas, 17th – 21st June 2019

Outline

Particle Flow Simulations Background

Background, prior research and findings.

Numerical Methods

BitCart, Dual-Mesh Approach, and AMR.

Simulations Results

Validation, and 2D/3D patricle flow simulations results.

Summary, Outlook, & Research Interest

Summary of presented research

4

Hypersonic Free Flight Disturbance Environment

Artist’s concepts of hypersonic cruise hardware Wave field in a hypersonic flow induced by disturbance sources

adapted from Zhong

• Understanding of the relevant physics is essential to

reduce design margins and systems uncertainties and, ultimately, guide the development of novel innovative designs

5

Hypersonic Free Flight Disturbance Environment

Artist’s concepts of hypersonic cruise hardware Wave field in a hypersonic flow induced by disturbance sources

adapted from Zhong

• Understanding of the relevant physics is essential to

reduce design margins and systems uncertainties and, ultimately, guide the development of novel innovative designs

• Disturbance environment and its effects on the

flow field need to be understood to provide accurate predictions

6

Hypersonic Free Flight Disturbance Environment

Artist’s concepts of hypersonic cruise hardware Wave field in a hypersonic flow induced by disturbance sources

adapted from Zhong

• Understanding of the relevant physics is essential to

reduce design margins and systems uncertainties and, ultimately, guide the development of novel innovative designs

• Disturbance environment and its effects on the

flow field need to be understood to provide accurate predictions

• Consider flow conditions at altitude of 15-45 km

(stratosphere) with a free-stream temperature range of 217 to 260 K and free-stream Mach numbers between 6-18

7

Hypersonic Free Flight Disturbance Environment

Artist’s concepts of hypersonic cruise hardware Wave field in a hypersonic flow induced by disturbance sources

adapted from Zhong

• Understanding of the relevant physics is essential to

reduce design margins and systems uncertainties and, ultimately, guide the development of novel innovative designs

• Disturbance environment and its effects on the

flow field need to be understood to provide accurate predictions

• Consider flow conditions at altitude of 15-45 km

(stratosphere) with a free-stream temperature range of 217 to 260 K and free-stream Mach numbers between 4-18

• Different types of particulates can be found

with ice clouds, a non-negligible amount of exhaust products from rockets, volcanic eruptions, terrestrial and cosmic dust, etc.

8

Hypersonic Free Flight Disturbance Environment

Artist’s concepts of hypersonic cruise hardware Wave field in a hypersonic flow induced by disturbance sources Research Objective: Provide physical insight into the interaction

• f the disturbance environment, in particular particulates, on

the flow field during realistic high-speed flight conditions

adapted from Zhong

• Understanding of the relevant physics is essential to

reduce design margins and systems uncertainties and, ultimately, guide the development of novel innovative designs

• Disturbance environment and its effects on the

flow field need to be understood to provide accurate predictions

• Consider flow conditions at altitude of 15-45 km

(stratosphere) with a free-stream temperature range of 217 to 260 K and free-stream Mach numbers between 6-18

• Different types of particulates can be found

with ice clouds, a non-negligible amount of exhaust products from rockets, volcanic eruptions, terrestrial and cosmic dust, etc.

9

Particle Properties in Atmosphere

• Particulates are inevitably present in the atmosphere as well

as in wind tunnels (unless careful cleaning technique), and they can be a major source of disturbance energy

• Properties and concentration of particles in the atmosphere

are documented in the literature (also see Hypersonic Flight In the Turbulent Stratosphere Research Team at UCB)

• Highly variable and seasonably dependent
• High concentration of particles can be obtained in ice clouds

(mostly in troposphere, regular crystalline shaped 𝒫(10- 1000µm)))

• Large amount of particulates are related to exhaust products

from rockets (𝒫(10µm))

• Another important source of particulates is volcanic

eruptions (𝒫(1-20µm)) Approximate size distributions for particles with different origins in the Earth’s middle atmosphere

(adjusted from Turco, data before 1992)

It is not a question of whether a flight vehicle encounters particles but rather how these particles affect the flow field around them!

10

Different mechanisms of how particles affect low and high-speed transition were summarized in Bushnell (1990):

Particle Flow Interaction Mechanisms

1) roughness generation via impacting or sticking to the surface,

11

Different mechanisms of how particles affect low and high-speed transition were summarized in Bushnell (1990):

Particle Flow Interaction Mechanisms

1) roughness generation via impacting or sticking to the surface, 2) vortex or vorticity shedding when particle is immersed in or external to the boundary layer,

12

Different mechanisms of how particles affect low and high-speed transition were summarized in Bushnell (1990):

Particle Flow Interaction Mechanisms

1) roughness generation via impacting or sticking to the surface, 2) vortex or vorticity shedding when particle is immersed in or external to the boundary layer, 3) boundary-layer mean shear can cause particle rotation and consequent fluid motions,

13

Different mechanisms of how particles affect low and high-speed transition were summarized in Bushnell (1990):

Particle Flow Interaction Mechanisms

1) roughness generation via impacting or sticking to the surface, 2) vortex or vorticity shedding when particle is immersed in or external to the boundary layer, 3) boundary-layer mean shear can cause particle rotation and consequent fluid motions, 4) ”reverse shocklets” can occur when particle passes through the vehicle-induced shock, and

14

Different mechanisms of how particles affect low and high-speed transition were summarized in Bushnell (1990):

Particle Flow Interaction Mechanisms

1) roughness generation via impacting or sticking to the surface, 2) vortex or vorticity shedding when particle is immersed in or external to the boundary layer, 3) boundary-layer mean shear can cause particle rotation and consequent fluid motions, 4) ”reverse shocklets” can occur when particle passes through the vehicle-induced shock, and 5) after particle impacts the surface it can rebound and dynamically interact with the bow shock induced by the vehicle causing the formation of jets and shear- layers. Ø Not a complete list very few fundamental studies have been conducted, especially for hypersonic flow.

15

Solver Overview & Simulation Approach

• Solving compressible Navier-Stokes equations with in-

house multi-physics solver BitCart (developed at UK)

§ Conservative FD scheme § Higher-order shock capturing (CWENO-6) for convective terms § 4th-order accurate treatment of viscous terms § Higher-order explicit and implicit time-discretization § Higher-order immersed boundary method (IBM) § Multi-species, gas chemistry, multi-phase, etc. § Fluid-structure interaction (FEM CSD solver) § Particle solver § Grid: generalized curvilinear, block-structured, adaptive mesh refinement (AMR) Cartesian, dual-mesh overset

• DNS of particle flows: solve nonlinear disturbance

equations with IBM, AMR, and dual-mesh approach

• Motivation was to develop method that has fidelity of

DNS but at a reduced computational cost. Simulation Domain Nonlinear Disturbance Flow Solver

16

AMR Dual-Mesh Approach

steady boundary layer flow disturbance flow stationary mesh AMR mesh

• AMR is a proven methodology for multi-scale problems with an extensive existing

mathematical and software knowledge base

• Higher-order accurate inter-level operators (implementation is similar to Kiris et al. (2018))
• Octree-based donor cell search algorithm for dual-mesh approach
• Sensitivity parameter 𝝌 controls mesh refinement/derefinement

𝜒 = 𝑛𝑏𝑦

()

*

+,- ()

*

,

(/

*

+,- (/

*

, … . based on tracking variable 𝜚3 𝒚, 𝑢, 𝑹′

• What is the best set of tracking variables? Compromise between efficiency vs. accuracy!

17

AMR Dual-Mesh Approach

Disturbance flow formulation of 3D compressible Navier-Stokes Equations Ø High grid resolution is only required locally, and temporal sub-cycling on the octree-based block-structured Cartesian mesh allows to efficiently simulate particles over time.

18

19

Comparison Against Standard DNS Approach

Sivasubramanian & Fasel 2016 Current AMR Approach

20

Sivasubramanian & Fasel 2016 Current AMR Approach

Comparison Against Standard DNS Approach

21

Sivasubramanian & Fasel 2016 Current AMR Approach

Comparison Against Standard DNS Approach

22

Particle Model

𝑒𝒚𝒒 𝑒𝑢 = 𝒘𝒒 𝑛𝑞 𝑒𝒘𝒒 𝑒𝑢 = <

=>? @A

𝑮𝒄,𝒐 + <

=>? @F

𝑮𝒕,𝒐

Kinematic equations: Newton’s second law: Different Positions Different Values of Collision Coefficient

a=0.7 & 0.6 a=0.76 a=1.0 xp: particle position vp: particle velocity Fb: body force Fb: surface force Ref=|vf - vp|dp/n is relative particle Reynolds number 𝜐I = 𝑆𝑓I𝑆I

L𝜍I/9

Sm: momentum source term Se: energy source term

• A. Particle-Source-In-Cell Method Simulation Approach

Surface force by Boiko (1997):

𝑮𝑻 = 𝑔

?

𝒘𝒈 − 𝒘𝒒 𝜐I − 1 𝜍I 𝜶𝑞 𝑔

? = 3

4 24 + 0.38𝑆𝑓𝑔 + 4 𝑆𝑓𝑔 1 + 𝑓𝑦𝑞 −0.43/𝑁

\ ].^_

Coupling with the fluid domain:

• B. Direct Particle Simulation via Immersed Boundary Method

where

Particle properties: V0=1180m/s, and Rp=10µm Particle properties: V0=1180m/s, and Rp=10µm

x x y y Ø Comparison in AIAA summer conference paper.

23

DNS of Particle Impingement for Mach 6 Flow

• Two-Step Simulation Approach:

1.) baseflow computation & 2.) AMR particle tracking simulation

• Particle flow simulation approach was initially tested in 2-D (or

axisymmetric) flows

• Initial 2-D simulations involve flows where second mode is the

most dominant instability mechanism

• 3-D simulations of are currently conducted and analysis will be

presented at AIAA summer meeting

Reproduced from Mack (1969)

Temporal growth rate

• vs. Me

24

Flow Visualization – Disturbance Pressure

• Mach 5.35 Boundary Layer Flow

After shock conditions: p∞=1297 Pa, r∞ =0.071 kg/m3, T∞ =63.9 K & Re=14.6·106m-1

• Particle properties:

rp=1000 kg/m3, Rp=5 µm Vp=[cos(7∘),sin(7∘)] 871 m/s xp=[0.11,0.004] m & Rep≈146 Surface Pressure Contours of Disturbance Pressure N-Factor (from LST)

25

Pressure Signature

• Disturbance pressure signal is

sampled at the wall via point probes,

• After performing an FFT on the

signals, the dominant and amplified wavenumbers/frequencies can be

• btained,
• The type of instability (first,

second, higher-modes, cross-flow etc) that is introduced can be identified,

• Calculating N-factors can be used

to predict transition location,

26

Pressure Signature

Pulse Disturbance − trigger second mode Particle Simulations

• Particles of size 10µm leads to non-dimensional pressure signature of 𝑞a

3/𝑞b = 𝒫 10cd − 10c]

• Pressure signature is highly dependent on flow conditions, particle properties, impingement location, etc.

Linear Nonlinear Particle collision location located upstream of neutral curve

27

Pressure Signature

Pulse Disturbance Particle Simulations

• Particle impingement at downstream location inside neutral curve for relevant frequencies
• Results for pulse and particle simulations are very different (due to initial disturbance level and receptivity)

Particle collision location located downstream of neutral curve

28

Pressure Signature

Pulse Disturbance Particle Simulations

• Particle impingement at downstream location inside neutral curve for relevant frequencies
• Results for pulse and particle simulations are very different (due to initial disturbance level and receptivity)

29

Effects of Particle Size

Particle with Dp=10µm

• Broad frequency spectrum introduced by particle collision (with low frequency peak)
• Particle collision location downstream of neutral curve,

low frequency Particle with Dp=50µm Particle with Dp=100µm

30

Effects of Particle Size

Pressure amplitude versus frequency for different particle sizes Estimate of excitation frequency

• As expected, larger particle size lead to vertical shift in amplitude curves
• In addition to higher disturbance amplitudes, a change in spectrum is observed
• Rough estimate for excitation frequency captures first peak in the spectrum (f ∼ vp/2d99)

31

Elastic vs. Inelastic Collision

Note: Dashed lines mark approximate path of particle.

Elastic Collision Inelastic Collision

32

3D Flat Plate BL Particle Flow Simulation

Adjusted from Churakov (2019)

• AMR is ideal candidate for

particulate induced transition simulations due to different scales that need to be resolved,

• Greater resolution needed

for computing particle trajectory than for resolving downstream wavepacket development,

• Grid levels can be removed

after particle collision detected.

3D Mach 5.35 Flat Plate Boundary Layer Flow

33

Simulation Setup: 7∘ Wedge, Mach 6

Adjusted from Churakov (2019)

3D Flat Plate BL Particle Flow Simulation

34

• Particulate collision with

boundary-layer upstream of neutral curve,

• Frequency-wavenumber

plots for various downstream locations,

• Wavepacket dominated by

second mode 2D instability,

• Validated against wall-

forcing simulations,

3D Flat Plate B-L Particle Flow Simulation

x=0.055 x=0.15 x=0.2 x=0.25

35

3D 14o Wedge Particle Flow Simulation

• M 4 wedge flow is first mode dominated
• Comparison with Chuvakhov (JFM 2019)

Flow conditions: b=474 Stability Diagram for 1st Mode (b=474)

36

3D 14o Wedge Particle Flow Simulation

37

3D 14o Wedge Particle Flow Simulation

38

3D 14 deg Wedge Particle Flow Simulation

39

Why Blue Waters HPC

• Transition simulations require a high grid resolution (both spatially and

temporally), not tractable on smaller computing clusters such as at UKy,

• Resources on Blue Waters enabled us to validate the AMR-WPT

methodology,

• It also enabled us to extend the method to studying particle induced

transition with AMR-WPT method,

• AMR refinement criteria,
• Running on 512-8191 procs, ramping up no. of procs, short queue times,
• Speed-up vs DNS,

40

Summary, Outlook & Research Interests

Summary and Outlook

• Efficient high-fidelity approach for particle-flow simulations has been

presented,

• Particle flow interactions are highly dependent on flow conditions, etc.
• Effects of particle properties (size, weight, and composition) will be further

investigated,

• Further analysis of particle flow simulations will be presented at the AIAA

Aviation,

41

Summary, Outlook & Research Interests

Thank you to everyone at Blue Waters for your help with my research and for hosting me at Sunriver, Oregon. Special mention for Brett! Any questions or comments?